Remote Activity Monitoring and Electronic Patient-Reported Outcomes Collection During Radiotherapy for Head and Neck Cancer: A Pilot Study

PURPOSE

Modern wearable devices provide objective and continuous activity data that could be leveraged to enhance cancer care. We prospectively studied the feasibility of monitoring physical activity using a commercial wearable device and collecting electronic patient-reported outcomes (ePROs) during radiotherapy (RT) for head and neck cancer (HNC).

METHODS

Patients planned for a course of external beam RT with curative intent for HNC were instructed to use a commercial fitness tracker throughout the RT course. During weekly clinic visits, physician-scored adverse events were recorded during using Common Terminology Criteria for Adverse Events version 4.0, and patients completed ePRO surveys using a clinic tablet or computer. Feasibility of activity monitoring was defined as collection of step data for at least 80% of the RT course for at least 80% of patients. Exploratory analyses described associations between step counts, ePROs, and clinical events.

RESULTS

Twenty-nine patients with HNC were enrolled and had analyzable data. Overall, step data were recorded on 70% of the days during patients' RT courses, and there were only 11 patients (38%) for whom step data were collected on at least 80% of days during RT. Mixed effects linear regression models demonstrated declines in daily step counts and worsening of most PROs during RT. Cox proportional hazards models revealed a potential association between high daily step counts and both reduced risk of feeding tube placement (hazard ratio [HR], 0.87 per 1,000 steps, P < .001) and reduced risk of hospitalization (HR, 0.60 per 1,000 steps, P < .001).

CONCLUSION

We did not achieve our feasibility end point, suggesting that rigorous workflows are required to achieve continuous activity monitoring during RT. Although limited by a modest sample size, our findings are consistent with previous reports indicating that wearable device data can help identify patients who are at risk for unplanned hospitalization.

INTRODUCTION

Curative treatment pathways for localized or locally advanced head and neck cancer (HNC) often include definitive or postoperative radiotherapy (RT), with or without concurrent chemotherapy. Acute treatment toxicities, which include dermatitis, mucositis, neutropenia, and inanition/dehydration, can be severe. Depending on the treatment setting and patient population, the likelihood of hospitalization during a course of RT for HNC can exceed 30%.^1-5 Factors that may be associated with elevated hospitalization risk include poor baseline nutritional status^2,3 and advanced age.¹ Admission during a course of RT for HNC can detract from treatment efficacy^6,7 and contribute to patient and caregiver distress.⁸

CONTEXT

• Key Objective

• Physical activity monitoring and electronic patient-reported outcome (ePRO) collection are promising tools that could enhance patient evaluation during cancer therapy and enable enhanced supportive care. This prospective trial evaluated the feasibility of monitoring physical activity using a commercial wearable device and collecting ePROs during radiotherapy for head and neck cancer (HNC).

• Knowledge Generated

• Prespecified goals regarding the completeness of activity data collection were not achieved. Still, we detected trends associating low recent step counts with increased risk for feeding tube placement and hospitalization, supporting further studies on this topic.

• Relevance

• Available evidence suggests that daily step count may serve as a dynamic predictor of adverse events during chemoradiotherapy for HNC. Rigorous workflows may be required to facilitate high-volume data collection.

Collecting patient-generated health data (PGHD), which includes patient-reported outcomes (PROs) and wearable device data, is a promising approach for objectively and quantitatively evaluating experiences of patients with cancer through treatment and in the survivorship setting.⁹ For patients with HNC treated with RT, recent studies suggest that collecting PGHD using modern technology is feasible¹⁰ and that simple measures (eg, daily step counts) can serve as predictors of short-term hospital admission risk.⁴

Here, we report results from a feasibility study collecting wearable device data and electronic patient-reported outcomes (ePROs) in adult patients receiving curative-intent RT for HNC.

METHODS

Patients and Study Devices

Adult patients with HNC who were planned for a course of fractionated RT with curative intent (with or without concurrent chemotherapy) were eligible for this observational, prospective, institutional review board-approved trial. Patients who were unable to read or speak English were not eligible for study enrollment. All study patients signed informed consent before enrollment. This trial is registered on ClinicalTrials.gov (identifier: NCT03489252).

The study schema is depicted in Figure 1. At the time of study registration, patients were given a commercially available fitness tracker (Fitbit Charge 3, Fitbit Inc, San Francisco, CA). Patients were asked to wear the device on the nondominant wrist starting on the day of RT simulation and until the final day of RT, with a use time goal of 22 hours per day. Patients were instructed to charge the devices, which had 5-day battery lives, at home. The research team synced patients' devices using a computer in the clinic once each week.

FIG 1.

Study schema. ePRO, electronic patient reported outcomes; OTV, on-treatment visit; RT, radiotherapy.

Study Assessments

Study patients had weekly physician visits during the RT course, which is standard for all patients receiving daily RT. Physician-scored adverse events were recorded during these visits using Common Terminology Criteria for Adverse Events (CTCAE) version 4.0. Patients also completed ePRO surveys during their weekly visits using a clinic tablet or computer. ePRO surveys included 18 PRO-CTCAE items relevant to the setting of RT for HNC, the Edmonton Symptom Assessment System (ESAS, a short palliative care questionnaire), questions regarding recent unplanned medical encounters, and questions about personal preferences and values regarding health monitoring.

Statistical Methods

The primary study objective was to determine feasibility of wearable device use during RT for HNC. We initially planned to define feasibility as daily device use of at least 19 hours on at least 80% of the days during the study period, defined as the day from RT simulation until RT completion. Because of difficulty obtaining device wear time data, we instead defined feasibility as collection of step data (>100 recorded steps¹¹) on at least 80% of the days during the study period. Feasibility data were reported using descriptive statistics and visualized using a Swimmer's plot. The intended sample size was 41 patients, which would provide 80% power to reject the null hypothesis that data collection is feasible in fewer than 60% of patients, assuming a true feasibility rate of at least 80%.

Secondary study objectives included evaluating acceptability of remote activity monitoring using a specialized survey, demonstrating feasibility of weekly ePRO collection in the study population, and exploring associations between physical activity data, ePROs, and unplanned medical visits (emergency department [ED] visits and/or hospitalizations).

For each patient and day, a 7-day step count average was defined as the average step count on that day and the 6 preceding days. The change in the 7-day step count average during treatment was modeled using mixed effects linear regression, with patient-specific random intercepts and a fixed effect for time in weeks since start of treatment. Missing steps were not imputed. A similar mixed effects model was used to estimate the changes in PRO-CTCAE grades¹² over time. The association between 7-day average step count and PRO grade was modeled using a mixed-effects model, with patient-specific random intercepts and slopes for time in weeks since start of treatment and fixed effects for average steps.

Unadjusted Cox proportional hazards regression models with time-varying predictors were used to explore the association between PRO grade and (1) ED visit or hospitalization during treatment and (2) percutaneous endoscopic gastrostomy or feeding tube during treatment. The hazard ratio (HR) of these events associated with a one-unit increase in PRO grade was calculated. Similar models were used to evaluate the association of 7-day average step count with both outcomes. Patients with events before start of treatment were excluded from these models. All statistical analyses were performed using SAS version 9.4.

RESULTS

Study Patients

Forty-one patients with HNC provided informed consent for study enrollment between February 2019 and August 2021. Because of technical issues, data from the first four patients were permanently lost. Data were not collected for several other patients because of withdrawal from the study (n = 5) or failure to meet all eligibility criteria (n = 2). Reasons for study withdrawal were severe treatment-related side effects and hospitalization (n = 2), discomfort with the wrist strap (n = 2), and onset of the COVID pandemic (n = 1), and study withdrawals took place between 1 and 4 weeks after study entry. Twenty-nine analyzable study patients are included in this report.

Patient characteristics are described in Table 1. The mean age at study entry was 61 years, and 83% of study patients were male. Most patients (72%) were treated with postoperative RT rather than definitive RT, and most (72%) received concurrent chemotherapy during RT. The most common HNC subsites were oropharynx (45%) and oral cavity (17%).

TABLE 1.

Patient and Treatment Characteristics

Feasibility of Remote Activity Monitoring

Step count data.

Overall, step data were recorded on 70% of the days during patients' RT courses. On a per-patient level, there were only 11 patients (38%) for whom step data were collected on at least 80% of days during the RT course. Table 1 includes comparisons of characteristics of patients for whom our feasibility end point was achieved versus other patients. Seven patients (24%) had missing data on the first day of RT, 17 patients (59%) had missing data on the last day of RT, and 14 patients (48%) had gaps in data collection during RT. These results are depicted graphically in Figure 2.

FIG 2.

Swimmer's plot depicting success of step data collection from the RT start date to the date of RT completion. RT, radiotherapy.

Baseline step count average, defined as the average step count over the first 7 days with available data, ranged from 1,453 to 15,158 (median 5,162). Regression models demonstrated an average daily step count decline of 418 (95% CI, 363 to 473) during each week of RT. Daily step counts over time are depicted in Figure 3.

FIG 3.

Daily step counts over time. Each line represents one patient. Step counts are smoothed using a moving average function for clarity. Red triangles denote hospital admissions, and red circles represent feeding tube placement. Step counts for patients who had a hospitalization or feeding tube placement event are shown using red lines, and step counts for other patients are shown using black lines.

PROs.

One hundred twenty-four PRO-CTCAE questionnaires were completed by 25 study patients. Among those patients, 18 (72%) reported at least one grade 3 adverse event. The most common grade 3 patient-reported adverse events were general pain (n = 13, 52%), decreased appetite (n = 11, 44%), and mouth/throat sores (n = 9, 36%).

Regression models demonstrated a statistically significant change over time in 8 of the 10 patient-reported symptoms evaluated using PRO-CTCAE. The greatest rates of change were observed for mouth and throat sores (+0.33 per week, 95% CI, 0.25 to 0.41, P < .001), decreased appetite (+0.25 per week, 95% CI, 0.18 to 0.33, P < .001), and general pain (+0.22 per week, 95% CI, 0.15 to 0.29, P < .001). Anxiety was the only patient-reported symptom that decreased during the treatment course (–0.06 per week, 95% CI, –0.12 to –0.03, P = .039). The only patient-reported symptoms for which increasing severity was significantly associated with low recent step counts were anxiety and sadness (Table 2).

TABLE 2.

Average Change of PRO-CTCAE Symptom Grades per Week During RT and Expected Change Associated With an Increase in Daily Step Count Average (over 7 days) of 1,000

Associations between PGHD and clinical events.

Three patients (10%) had ED visits during RT, and all three were subsequently hospitalized. Nine patients (31%) had feeding tubes placed during the RT course.

Cox proportional hazards models for time-varying predictors (Table 3) revealed a potential association between high average daily step counts and reduced risk of hospitalization (HR, 0.60 per 1,000 steps, 95% CI, 0.49 to 0.73, P < .001), indicating a 40% reduction in the risk of hospitalization for every increase of 1,000 in average steps taken during the previous 7 days. There was also evidence of a potential association between higher average daily step counts and reduced risk of feeding tube placement (HR, 0.83 per 1,000 steps, 95% CI, 0.77 to 0.90, P < .001). The only PRO for which there was evidence of association with hospitalization risk was shortness of breath (HR, 11.54 per point, 95% CI, 1.16 to 114.90, P = .042). PROs were not significantly associated with risk of feeding tube placement.

TABLE 3.

Associations Between Step Count Data and Patient-Reported Outcomes and Risk of Hospitalization or Feeding Tube Placement

DISCUSSION

We performed a pilot study testing the feasibility of collecting wearable device data and ePROs in adult patients receiving curative-intent RT for HNC. Even after excluding patients who consented for this study and subsequently withdrew, we did not achieve our prespecified goal with respect to completeness of wearable device data. However, we gathered a large quantity of novel data in a vulnerable patient population. We found no strong evidence that daily step counts can be predicted using PROs, and we found that low daily step counts may be a risk factor for feeding tube placement, ED visit, and hospitalization during RT.

Missing step data was most common at the end of patients' RT courses. Patients apparently grew tired of wearing and/or charging their devices over time, which could be a result of accumulating treatment toxicities and/or waning interest in devices that were not providing tangible user benefits. Reviewing recent step data with patients during clinic visits, which was not done in this study, could increase patients' interest in continued device use. Some patients had missing data at the start of therapy, suggesting that additional education of study patients and real-time monitoring for data completeness could have improved data quality. Alternative data collection strategies, such as syncing patients' devices to their personal mobile phones for passive data collection or use of wireless receivers to collect device data, could facilitate more complete data collection. Wearable devices that do not require frequent charging might be more practical for some patients. We were not able to collect adequate data to present of all of our planned exploratory analyses (eg, assessing patient perceptions of activity monitoring and data sharing) because of clinic staffing shortages during the peaks of the COVID pandemic. Future study of patient perceptions may aid in addressing barriers to successful implementation of wearable devices.

Previous studies testing the feasibility of wearable device use during chemotherapy or RT for solid tumors have yielded mixed results. Selected trials are summarized in the Appendix Table A1. Our interpretation of these results is that most patients are willing to use wearable devices during a course of daily RT, but vigilance from the clinic team is needed to identify cases of missing data in real-time and take corrective actions. Wearable devices still hold promise as tools in interventional trials and could reasonably be incorporated as routine patient assessments in the future.

Our ability to evaluate associations between PGHD and clinical events was constrained by the small sample size and low number of events. Despite our limited sample size and few hospitalization events, we found a trend suggesting an association between low recent step counts and increased hospitalization risk. This is consistent with results from Montefiore trials.^4,10 This association was not seen in a trial performed by the UT Southwestern group.¹³ However, 13 of 15 hospitalizations in the UT Southwestern study were planned for feeding tube placement. This demonstrates the complexity of using hospital admission as a study end point and suggests that aggressive supportive care interventions can reduce the risk of unplanned hospital encounters. Of note, practice patterns regarding feeding tube placement during HNC RT vary widely across institutions.¹⁴ Future studies could explore if wearable device data could supplement standard measures (eg, weight loss) as triggers for feeding tube requirement.

Interestingly, the only PROs that were statistically significantly associated with recent step counts were anxiety and sadness. In both cases, more severe symptoms were associated with lower step counts. Associations between low mood and low step counts have been reported in several studies of healthy adults.^15,16 Additional studies of physical inactivity as an indicator of psychiatric issues during cancer therapy are warranted.

In conclusion, we did not achieve our feasibility end point, suggesting that rigorous workflows are required to achieve continuous activity monitoring during RT for HNC. Although limited by a modest sample size and small number of events, our findings are consistent with previous reports indicating that wearable device data can help identify patients who are at risk for unplanned hospital visits.

APPENDIX

TABLE A1.

Selected Trials Testing the Feasibility of Wearable Device Use During Chemotherapy or RT for Solid Tumors

AUTHOR CONTRIBUTIONS

Conception and design: Voichita Bar-Ad, Benjamin Leiby, Gerard Hoeltzel, Camilo Henao, Adam P. Dicker

Financial support: Adam P. Dicker

Administrative support: Camilo Henao, Adam P. Dicker

Provision of study materials or patients: Gerard Hoeltzel

Collection and assembly of data: Nitin Ohri, Voichita Bar-Ad, Christian Fernandez, Gerard Hoeltzel, Anna Sung, Nida Zubair, Camilo Henao, Adam P. Dicker

Data analysis and interpretation: Nitin Ohri, Voichita Bar-Ad, Christian Fernandez, Christine Rakowski, Benjamin Leiby, Gerard Hoeltzel, Anna Sung, Camilo Henao, Adam P. Dicker

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/cci/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).